Digital Imaging: Implications for Image Quality and Radiation Dose

Overview

• Digital imaging has fundamentally transformed radiological practice, replacing analogue screen-film systems with flexible, post-processable, electronically stored images.
• This transition has brought profound advantages in image quality optimisation and dose management, but also introduced new pitfalls, most notably the decoupling of exposure level from apparent image quality.
• Understanding how digital imaging techniques affect contrast, spatial resolution, noise, and dose is central to the RANZCR Part 1 examination and to safe, effective clinical practice.

Principles of Digital Image Formation

From Analogue to Digital

• In screen-film radiography, the image receptor imposed a fixed relationship between exposure and optical density.
• Underexposure produced a pale film; overexposure produced a dark, unusable film, an inherent safety feature that naturally limited dose creep.
• Digital detectors, whether computed radiography (CR) photostimulable phosphor plates, or direct/indirect flat-panel detectors (FPD), convert incident X-ray energy into electrical signals that are digitised into a matrix of discrete pixel values.
• The resulting image can be post-processed, windowed, and distributed electronically, completely independent of the original exposure level.
• Critically, overexposure does not degrade the displayed image and may go unrecognised by the operator.

Key Digital Detector Parameters

Image Quality in Digital Systems

Spatial Resolution

Spatial resolution is limited by the pixel pitch (physical size of each detector element) and by the modulation transfer function (MTF) of the entire imaging chain. The Nyquist theorem defines the maximum spatial frequency that can be faithfully sampled:

$$f_{\text{Nyquist}} = \frac{1}{2 \,\Delta x}$$

where $\Delta x$ is the pixel pitch in millimetres. Frequencies above this limit are aliased, creating false patterns in the image.

• Digital systems generally have slightly lower spatial resolution than analogue screen-film systems, but this is offset by the ability to zoom, pan, and window the image post-acquisition.
• Dedicated high-resolution applications (e.g. mammography, skeletal extremity imaging) use smaller pixel pitches and high-resolution detector elements; mammography specifically requires small focal spots and very high-resolution detectors to depict microcalcifications.

Although digital images have lower intrinsic spatial resolution than their analogue counterparts, the ability to adjust contrast and windows and to view images remotely via PACS has driven widespread adoption.

Contrast Resolution and Windowing

• One of the greatest advantages of digital imaging is adjustable contrast.
• The look-up table (LUT) maps raw pixel values to display brightness levels.

Window width (WW) controls the range of pixel values displayed across the full greyscale; window level (WL) sets the midpoint:

$$\text{Displayed contrast} \propto \frac{1}{\text{Window Width}}$$

• Narrowing the window width amplifies contrast differences between tissues but sacrifices the visible dynamic range.
• This allows the same raw dataset to be interrogated at different contrast settings (e.g. lung windows WW ~1500 HU vs. mediastinal windows WW ~400 HU in CT).
• LUT selection can have a profound effect on clinical performance and procedural radiation utilisation.
• Suboptimal display during a fluoroscopic procedure can lead operators to extend fluoroscopy time, increase dose rates, or substitute higher-dose fluorographic acquisitions simply to achieve the image quality needed, effects that cannot be detected by standard quality testing of dose-rate settings or technical monitor performance values.
• Most fluoroscopic systems provide several LUTs tailored to specific clinical tasks (e.g. abdominal DSA), selectable at the console.

By contrast, screen-film systems had a fixed sigmoid characteristic curve; optimal contrast was achievable only within a narrow latitude of exposure.

Noise and Signal-to-Noise Ratio

Quantum noise (quantum mottle) arises from the statistical variation in the number of X-ray photons absorbed per pixel. It follows Poisson statistics:

$$\sigma_N = \sqrt{N}$$

where $N$ is the mean number of photons detected per pixel. The SNR therefore scales as:

$$\text{SNR} = \frac{N}{\sqrt{N}} = \sqrt{N}$$

• To double the SNR, the dose must be quadrupled ($\propto N$).
• This fundamental quantum noise constraint applies to all X-ray-based digital imaging.

Contrast-to-noise ratio (CNR) combines contrast and noise:

$$\text{CNR} = \frac{|S_1 - S_2|}{\sigma_{\text{noise}}}$$

where $S_1$ and $S_2$ are mean signal values in two regions. CNR determines lesion detectability and is the primary driver of dose requirements in practice. Image quality measures based on SNR and CNR serve as surrogates for diagnostic accuracy.

Image Processing Techniques and Their Quality/Dose Implications

Spatial Blurring

• Blurring algorithms replace each pixel value with a weighted average of neighbouring pixels (e.g. Gaussian kernel).
• The number of neighbours and their weighting are both programmable.
• This suppresses high-frequency quantum noise at the cost of reduced spatial resolution, a larger kernel yields more noise suppression but greater loss of fine detail.

Fluorographic (spot) images typically have less noise than fluoroscopic images; consequently, smaller blur kernels are clinically acceptable for fluorography, which is why fluorographic images appear sharper than fluoroscopic images of the same anatomy.

Dose implication: Effective blurring allows acceptable image quality at lower doses, provided fine structural detail is not clinically critical.

Unsharp Masking and Edge Enhancement

Unsharp masking subtracts a blurred version of the image from the original to create a sharpened output:

$$I_{\text{sharp}} = I_{\text{original}} + k \times (I_{\text{original}} - I_{\text{blurred}})$$

where $k$ controls the degree of enhancement. This improves apparent sharpness of edges (Mach band effect), aiding detection of linear structures, cortical bone margins, and vessel walls. The process is analogous to edge-enhancement interactions in the retina at a neural level.

• Dose implication: Edge enhancement amplifies high-frequency content including noise.
• At lower doses (noisier images), unsharp masking may amplify noise artefacts rather than true edges, and overuse can create artefactual edge ringing or obscure subtle low-contrast lesions.

Local and Adaptive Image Processing

• Modern systems apply different processing algorithms to different spatial regions of the same image, combining a-priori knowledge of procedure type with real-time feature extraction from the image sequence.
• For example, within a single fluoroscopic frame, bone edges may be enhanced while soft tissue regions are smoothed.
• Evaluating and optimising fluoroscopic image processing using technical tests alone is insufficient, because patient and operator factors significantly influence clinical performance.

Dose implication: Adaptive algorithms can achieve diagnostically adequate image quality at lower overall dose by directing resolution enhancement where needed and suppressing noise elsewhere.

Temporal Averaging (Recursive Filtering)

• In fluoroscopy, successive frames are averaged to reduce noise.
• If $n$ frames are averaged, noise reduces by $\sqrt{n}$.
• However, rapidly moving structures (e.g. coronary arteries, bowel peristalsis) produce motion blurring in heavily averaged images.
• Dose implication: Temporal averaging is a key dose-reduction strategy in fluoroscopy.
• It is only appropriate when motion is limited.

Pulsed Fluoroscopy

• The X-ray generator is switched on and off to supply discrete pulses rather than a continuous beam.
• Continuous fluoroscopy is equivalent to approximately 25-30 pulses/s (each pulse 2-20 ms); the pulse rate can be reduced to 15, 7.5, 3.75 pulses/s, or lower.
• There is no perceived flickering because each image is held until the next pulse is generated, maintaining frame rate above the critical flicker frequency.

Dose implication: Reducing pulse rate from 25 fps to 7.5 fps reduces dose rate by approximately 70% while maintaining adequate temporal resolution for many procedures.

Digital Subtraction Angiography (DSA)

• DSA acquires a pre-contrast mask image and subtracts it from subsequent contrast-filled images, eliminating fixed background structures to reveal vessels or other moving/changing structures.
• After subtraction, window and level are greatly widened to improve vessel visibility, which simultaneously increases the visibility of quantum noise in the subtracted image.

Noise does not subtract, it adds in quadrature:

$$\sigma_{\text{DSA}} = \sqrt{\sigma_{\text{mask}}^2 + \sigma_{\text{live}}^2} \approx \sqrt{2}\,\sigma_{\text{single frame}}$$

• To achieve acceptable CNR in the subtracted image, both mask and live frames must be acquired at substantially higher dose than equivalent non-subtracted fluorographic frames.
• Nominally, up to ten times the dose per frame is required for a single DSA frame compared to a comparable unsubtracted (DA) fluorographic image.
• Dose implication: DSA is intrinsically higher dose per frame, but DSA series are acquired at low frame rates (programmable on a run-by-run basis to match contrast flow dynamics), so total radiation dose remains acceptable.
• Using the lowest practicable frame rate reduces total dose.
• Misregistration artefact from patient motion between mask and live frames can be corrected by remasking or pixel-shift algorithms.

Dual-Energy Subtraction (DES)

• DES acquires two rapid sequential exposures at different kVp (e.g. 60 kV and 120 kV for chest), exploiting differences in photoelectric attenuation between bone and soft tissue.
• Mathematical combination produces a standard PA image, a bone-subtracted soft-tissue image, and a bone-only image from a single acquisition pair.

Clinical applications include improved detection of lung nodules, identification of calcification within granulomas, visualisation of bone islands and healing rib fractures, and enhanced delineation of indwelling lines and catheters.

• Dose implication: Two exposures are required per acquisition, increasing total dose.
• The additional diagnostic information may justify this dose premium in appropriate clinical contexts.

Dose Implications of Digital Imaging: The Problem of Dose Creep

Decoupling of Exposure from Displayed Image Brightness

• In screen-film radiography, overexposure produced an unacceptably dark film, a visible, inherent deterrent to excessive exposure.
• In digital imaging, the displayed image brightness is determined by post-processing, entirely independent of actual detector exposure.
• An image acquired at twice the necessary dose will appear identical after LUT adjustment.
• This creates dose creep: a gradual, undetected escalation of patient dose, driven by preference for lower-noise images without clinical need.
• Routine monitoring of detector exposures, particularly in mobile radiography where AEC is commonly unavailable, is therefore a necessary quality assurance measure.

Exposure Index and Deviation Index (IEC 62494-1)

To address dose creep, the international standard IEC 62494-1 introduced the Exposure Index (EI) and Deviation Index (DI):

$$\text{DI} = 10 \log_{10}\!\left(\frac{\text{EI}}{\text{EI}_T}\right)$$

where $\text{EI}_T$ is the target exposure index for a given examination type. Negative DI values indicate underexposure; positive values indicate overexposure; DI = 0 indicates optimal exposure.

This metric provides immediate feedback to the technologist and forms a cornerstone of quality assurance in digital radiography.

Dose Optimisation Principles (ALARA/ALARP)

The principle of optimisation (ALARA/ALARP) requires that any exposure be as low as reasonably achievable while maintaining adequate image quality for the clinical task. Digital imaging supports this by enabling:

• Lower dose acquisition with noise suppression via post-processing
• Fewer repeated exposures due to wide dynamic range
• Quantitative exposure monitoring via EI/DI feedback
• Iterative reconstruction in CT to permit tube current reduction while maintaining diagnostic quality
• Higher DQE detectors reducing dose for equivalent SNR
• However, these advantages are only realised when appropriate protocols are implemented and monitored.
• Dose limits do not apply to patients undergoing medical exposures; patient doses are instead managed through Diagnostic Reference Levels (DRLs).

Diagnostic Reference Levels and Achievable Doses

• A dose metric exceeding the DRL should be investigated but is not automatically considered excessive.
• The ALARA principle mandates that image quality be adequate, not maximal, for the clinical task.

Specific Digital Imaging Modalities: Quality and Dose Considerations

Computed Radiography (CR)

• CR uses photostimulable phosphor plates; the stored latent image is released by laser scanning.
• CR has lower DQE than flat-panel detectors, is more susceptible to noise at low exposures, and has slower workflow.
• It has been largely superseded by direct-readout DR systems.

Direct Radiography / Flat-Panel Detectors (FPD)

FPDs use either:

• Indirect conversion: scintillator (e.g. caesium iodide) + amorphous silicon TFT array
• Direct conversion: amorphous selenium (photoconductor directly generates charge)
• FPDs have higher DQE, faster readout, and immediate image preview.
• Their wide dynamic range and high DQE allow lower patient doses compared to CR or screen-film for equivalent image quality.
• Post-processing capabilities allow simultaneous optimisation of bone and soft tissue from a single exposure, reducing the need for repeat acquisitions.

Fluoroscopy and Digital Fluorography

Digital spot (fluorographic) images use a 1024 × 1024 acquisition matrix with 10-bit greyscale and high mA, resulting in reduced quantum mottle relative to fluoroscopy.

CT: Dose Reduction Techniques

• Dose implication: Iterative reconstruction allows significant dose reduction in low-dose protocols (e.g. renal stone CT, paediatric CT, CT colonography) without diagnostic compromise.
• Awareness of altered noise texture at high IR levels (smooth, "waxy" appearance) is important, as this may mask or mimic pathology.

Artefacts Specific to Digital Imaging

Paediatric Considerations

Children are more radiosensitive than adults (longer remaining lifespan, more rapidly dividing cells) and smaller body habitus means adult exposure parameters yield disproportionately higher doses. Digital imaging benefits paediatric practice by:

• Wide dynamic range reducing repeat exposures
• Post-processing allowing simultaneous optimisation of bone and soft tissue on a single exposure
• High-kV/low-mAs protocols (dose-minimisation curves) reducing photoelectric absorption and skin dose
• CR and FPD technology providing greater efficiency in converting incident X-ray energy into image signal compared to screen-film
• Despite these advantages, the flexibility of digital detectors creates a risk of unrecognised overexposure if exposure factors are not carefully set and monitored.
• Unlike film-based techniques, overexposure occurs without adverse effect on displayed image quality.

Summary: Digital vs. Screen-Film

Clinical and Practical Implications for the Radiologist

1. Display optimisation matters: Suboptimal LUT or window settings during fluoroscopy can lead operators to extend fluoroscopy time, increase dose rates, or substitute fluorography for fluoroscopy, escalating dose in ways undetectable by standard technical quality testing.
2. DI monitoring is essential: Regular review of Deviation Index values identifies systematic overexposure before it becomes entrenched practice.
3. Processing cannot recover information destroyed by quantum noise: Image processing compensates for noise but cannot restore signal lost at very low doses. There is a minimum acceptable dose floor for every examination type.
4. Adequate, not perfect: The goal of digital optimisation is images that are diagnostically sufficient for the clinical task, not the lowest-noise, highest-resolution image achievable.
5. Iterative reconstruction in CT enables meaningful dose reductions but requires awareness of altered noise texture at high reconstruction strengths.
6. Frame rate selection in DSA and fluoroscopy should be matched to the temporal requirements of contrast flow and clinical anatomy, using the lowest practicable frame rate reduces total dose without sacrificing diagnostic information.